JPH06110740A - Method and means for measuring channel-using time - Google Patents

Abstract

PURPOSE: To provide a method for measuring the use time of an input/output channel that one of operating systems(OS) in a computer complex(CEC) uses. CONSTITUTION: A channel measuring mechanism(CMF) is provided for input/ output subsystem hardware and microcodes for the respective OSs in an input/ output subsystem. Resources for a logical CMF includes partial use of an input/ output processor for storing the identifier of an assigned OS, partial use of a channel processor which controls a channel selected by the assigned OS in a measurement period, a local storage area for respective input/output processors, and an OS storage area needed to communicate measurement data from the CMF to the OS.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention provides a channel path measurement mechanism within a channel subsystem and a computer that supports input / output resource sharing as described in US patent application No. 898867 filed on the same date as the present application. Complex (CE
Allows the program to store usage information for operations on both shared and non-shared I / O channel paths of C). The entire contents of the following US patent applications filed on the same date as the present application are hereby incorporated by reference. U.S. Patent Application No. 898867,
Nos. 898977 and 898875.

The following US patent applications also form a part of this specification by referring to the contents thereof. U.S. Patent Applications Nos. 07 / 444,190 and 07 /
No. 754813, No. 07/676603, No. 07
/ 755246, and No. 07/693997.

[0003]

BACKGROUND OF THE INVENTION When measuring the utilization of input / output resources, the prior art offers various solutions and techniques. The present invention proposes a new method and means which overcomes most of the drawbacks of these prior art solutions.

In order to evaluate the prior art and compare it with the proposed invention, the requirements for measuring the activity of input / output resources are summarized as follows.

-Measure the total time that the I / O resource is working on behalf of the measuring OS. Give this measurement at any time interval of duration from a few seconds to a few hours. -Allow any start and end points for any measurement interval. -Provides information that enables measurement of multiple overlapping intervals that use input / output resources shared by multiple OSs. For example, O
While S A measures I / O resource activity between 8:00 and 9:00, other OS Bs have
It must be possible to measure activity using the same resources between 0 and 9:05. The two OSs must be given measurements that do not indicate that the other OS is using shared I / O resources. -Measurements shall be accurate within user specified limits. -The overhead in the CEC must be minimal in making the measurement. -The programming of the measurement program should be simple.

[0006] One of the existing methods is to use a statistical technique for measuring channel utilization to periodically check the busy state of the channel and estimate the total use time based on the statistical technique. . The advantage of the statistical technique is that no special activity counter is needed for the channel. The disadvantages are:

There is no way to correlate the channel state to the OS where the I / O activity is done by the CEC channels of multiple OSs. -Frequent sampling brings significant overhead in terms of CPU and channel subsystem utilization. -In addition to sampling activities, the OS needs to frequently transform these statistical samples into data that can be used by numerous other programs running on the OS to make the measurements meaningful. -To minimize CPU overhead,
The accuracy of such measurements becomes questionable if the sampling frequency is reduced. -Sampling-based measurements are inaccurate compared to continuous measurements. − Due to various constraints imposed by system design,
Sampling may be biased.

Another conventional sampling technique is to sample the channel utilization within the channel itself, estimate the total utilization of the channel, and periodically store the information in a memory available to the measurement OS. The advantage of this method is OS
That is, there is no need to frequently sample the channel or transform the sampling result. The drawback is the inability to extend such a technique to associate the activity of each channel with its initiating OS and allocate the observed usage time to each OS.

A third prior art technique (the technique currently used to drive controller queue related activities).
Maintains counters and accumulators in the channel subsystem and only when requested by the OS
It stores information. The advantage of this method is that the CEC overhead is minimized when the measurement information is not frequently needed.

[0010]

The drawback is that when information is obtained in a very short period of time, it frequently executes instructions used by the OS to request information from the channel subsystem, as well as the CPU and channel subsystem. And the associated communication required between them results in excessive CPU and channel subsystem overhead.

The prior art device connect time measurement causes the channel subsystem to accumulate a period of connect time for each I / O device and increment the designated storage location at the completion of each I / O operation. The advantage lies in the device-level granularity and high precision. However, if total channel utilization measurements are needed, all device-level measurements have to be aggregated for the devices configured for the channel to determine total channel utilization. Since it does not exist, the advantage of device level granularity is a drawback.

Furthermore, prior art device level channel measurements do not provide individual utilization measurements for each of the plurality of channels that can comprise the device. That is,
Since multiple channels may be involved in each I / O operation to each device, the desired information regarding channel utilization cannot be obtained by adding device-related information. Moreover, storing the information from all participating channels at the completion of all I / O operations makes the overhead to the channel subsystem unacceptable.

Finally, all prior art I / O resource measurement mechanisms selectively select I / O resources shared by individual OSs in a CEO where I / O channel resources are dynamically shared among multiple OSs. It does not have a function to measure the usage rate.

The following US patent applications form the background of the present invention. U.S. Patent Applications Nos. 07 / 444,190, 07 / 754,813, 07 / 676,603, 07 / 755,246, and 07 / 693,997.
issue.

[0015]

The solution provided by the invention is a combination of the following:

-Each channel shares a plurality of OSs
Accurate accumulation of individual usage time for each of the. -Each channel periodically stores this information in an area available to the I / O subsystem. The frequency of this store operation is low enough not to cause any significant overhead. -When storing cumulative measurements, the channel processor also stores the time stamp associated with the measurements. This time stamp allows the delay check of the measurement by the OS to be adjusted to the desired accuracy, reflecting the exact channel utilization at the closest time. Without these time stamps, the accuracy of the measurements will be greatly affected by the frequency with which the measurement data is stored. If the storage frequency is low, the overhead is low and the accuracy is low. Frequent storage increases overhead but improves accuracy. Timestamps allow for both infrequent storage and high accuracy. -The Channel Subsystem periodically moves the information stored by the Channel Processor to the OS storage. The time stamp is preserved, so the delay in this regular movement is also taken into account as described above.

The present invention provides a Channel Measurement Facility (CMF) that can be used by the CEC control program. This is particularly advantageous when used by any one of a plurality of OSs in logical partitions having different CECs. For each OS, each CMF is independent of whether or not the channel is switched between I / O devices during the measurement time, and whether or not the channel is switched between multiple OSs during the measurement time. Support continuous time measurement of usage for any channel, including shared and non-shared channels operating on any OS. In the case of shared channels, the operation of the present invention relies on the use of the invention described in U.S. Pat. ), A new method and means for controlling I / O resources that allow multiple operating systems (OS) running in CEC to be dynamically shared (operate non-shared) are described.

[0018] US Patent Application No. 878867 enhances I / O channel connectivity to multiple operating systems running in different logical resource partitions of a computer complex (CEC) to provide channel connectivity between OSs within a CEC. The method of sharing is described. The application assigns an image identifier (I
ID) is disclosed. In the CEC channel subsystem, each shared I / O is provided with a shared set of control blocks (CBs), each CB being assigned to a respective IID of one of the OSs operating in the CEC. (It is arranged by this). Each CB in the shared set gives a different image to each OS of the same I / O resource. Since different CB images are independently set in different states by I / O operations where the resources are for different OSs, the OS independently operates on the same I / O resource regardless of the order in which the OS uses the resources. can do. The IID used when the OS executes the I / O request is transmitted to the I / O controller used to access the requested I / O device and is controlled as part of the logical path connecting to the control unit. Stored by the unit and used later by the control unit in response to the requesting OS for I / O requests.

The present invention relates to the measurement of the use time when each channel is used for a specific OS regardless of whether the use of the channel is shared by a plurality of OSs during the measurement period or not. Without using any sampling means to measure the channel usage time, the present invention directly and accurately measures the channel usage time. For this reason, the channel measurement facility offloads the direct measurement function from the requesting CPU to the channel processor operating asynchronously in the I / O channel subsystem, thereby asynchronously with the CPU requesting the measurement. Operate.
Upon completion of asynchronous measurement operations, these operations are periodically updated to the OS.

The preferred embodiment of the present invention is independent of whether the control program is running under a multi-OS CEC or a single-OS CEC OS, and when the channel is run by another OS during the channel measurement period. O, whether or not it is shared
Provides a new CPU instruction used by the OS control program to request a channel measurement operation from S. This CPU instruction is used to initiate a measurement operation by the channel measurement mechanism for the requesting OS. Also, this CPU
The instruction is used to stop the measurement operation of the channel measurement mechanism for the requesting OS. In addition, this CPU instruction is used to test the state of the channel measurement mechanism to the requesting OS, such as whether the channel measurement mechanism is active in a measurement operation, is in a stopped state, or is in an error state.

The time measurement itself is saved to the channel processor which directly controls the channel itself. The result is that the CPU overhead with respect to channel usage time remains constant and does not increase when the sampling frequency of channel usage increases as in the prior art.

For efficient direct time measurement, a channel measurement mechanism for each requesting OS is placed in the hardware / microcode of the CEC channel subsystem, which starts and stops each channel. And an asynchronously operating processor that controls the operation of the channel, including testing the status of the channel measurement mechanism.

An asynchronously operating processor within the CEC channel subsystem is an input / output processor (IO) for passing the I / O work requested by the CPU to the channel.
P) and may also include a channel processor that receives work requests from the IOP and directly controls the operation of each channel.
Each channel processor can be shared among multiple channels, or each channel processor can be dedicated to a particular channel. The preferred embodiment of the present invention uses the latter method in which each channel has its own dedicated channel processor which continuously controls and monitors the usage of that channel. To do.

Placing the IOP and channel processor in the CEC microcode to start, stop, and test the channel measurement mechanism for each OS, eliminating the need for the CPU to have this microcode. Is preferred. This is because the CPU already has a heavy load due to the microcode for the very many functions that the CPU requires.

As a result, the CPU instructions provided by the present invention only communicate start, stop or test functions to the channel subsystem IOPs and channel processors, and also return completion information to the OS. . The channel subsystem then asynchronously and periodically, without incurring CPU overhead, all configured for each OS, regardless of whether any of the channels are shared by other OSs. It provides a direct and continuous measurement of the usage / non-use status of the channels.

CPU instructions are referred to as "set channel measurement" (SCM) instructions in the preferred embodiment described herein. If the CEC is operating with shared channels,
The SCM instruction causes the channel subsystem to measure only the portion of the requesting OS that is using the shared channel, or, in the case of an OS, only the portion of the requesting program operating under that OS. To measure. It is not reported to the requesting OS that each of the other shared OSs is using the shared channel proportionally, but these other O
It is possible to report to each of S.

The ability of the CEC to work with both shared and non-shared I / O resources can be obtained when the CEC is using either a microcoded hypervisor or a software hypervisor.
Multiple O's operating in the CEC with any of these types of hypervisors using SMC instructions.
It is also possible to start or stop the measurement for any one of S. C with multiple OSes using SCM instructions with any of these hypervisors
The status of the channel measurement mechanism can be tested by the designated OS in the EC.

When active in the CEC's channel subsystem, the channel measurement facility continuously stores channel measurement information in the CEC's channel processor. However, CEC resources are affected by the number of times measurement data is delivered to the OS requesting this data. Experience has shown that data does not have to be continuously transmitted to the OS, but periodically over a relatively long period of time, or using a special CPU instruction to obtain updated data for the OS. Has been shown to be sufficient. These two are included in the present invention. Regular reporting has the advantage that it does not burden the OS with the software that needs to be called periodically to perform the update process, and the additional instructions in its repertoire cause the CPU to do so. There is no burden on the CPU. The disadvantage of regular reporting is that it is not guaranteed to be timely for long periods of time, which is the shortest time for measuring channel uptime that does not need to be reported very timely. .

It has been found that intervals of seconds can be effectively used for regular reporting. For example, a 4 second interval may be used to update the channel usage information provided to the OS.

It has been found preferable to have the channel processors report their channel usage time to their respective OS, rather than first to the control blocks that have access only to the CEC hardware / microcode. For example, the channel processor could report consecutive measurements of channel usage time to each CMF at 2 second intervals, which could be either to the channel processor in the CEC or to the hardware / microcode. Not a significant burden. And every 4 seconds, the updated measurements in the hardware / microcode control block are periodically copied to the control block in the storage allocated by the OS in the CEC, and the information is available to each OS. To do so.

In addition, the OS can be enabled to obtain channel measurement data for other selected OS's in the CEC. This operation is subject to all security and authority control between OSs in the CEC. This is controlled by the OS issuing a measurement instruction that identifies the other OS in the CEC where the measurement is made. The OS can issue multiple measurement commands to simultaneously perform measurements on multiple other different OSs.

In addition, the OS can obtain channel path measurement data for other selected OSs.
This operation is subject to all security and authority control between OSs in the CEC. This is another O
It is controlled by the OS that issues measurement instructions that identify S. The OS can issue multiple measurement commands to request measurements to other multiple different OSs simultaneously.

Measurement Control by Other OS The OS can control the measurement of channel utilization with respect to the other OS. To do this, the OS sets the "OS select IID" field in the associated command request block (CRQB) for the IID of the other OS for which the measurement is made, and then issues a "set channel measurement" (SCM) command. To do. See FIG.

Next, in CRQB, "OS selection II" is used instead of IID in the "OS image ID" field.
The IID in the "D" field is used to specify the CMF that is the subject of this SCM instruction.

When measuring the OS, "OS image I
If the "D" field is used to specify the requesting OS, the value of the "OS Select IID" field is set to zero.

In accordance with the selection of an IID other than the IID of the requesting OS, the command processing by the CMF, the data collection, the data movement, and the error processing are executed by the OS in the CM associated
It is done in the same manner as when starting F, and all the steps described above are performed except for the following.

CMF for a given OS is stopped, SC
If M starts, a command is issued for the data associated with this CMF, then a measurement for this CMF is initiated in the same manner as the command was issued from the OS associated with the CMF. The requesting OS is then registered by the IOP as a recipient of CMF data for the current measurement.

The IOP registers all OS's that have an active CMF. The IOP status block of FIG. 7 shows the registration of options for each OS to receive CMF data. The IOP status block (FIG. 7) can be duplicated for this purpose. The IOP maintains a separate status block for each OS it collects data from, which is
List all OSs that are registered to receive OSs for. This registration contains the CMB address that the requesting OS wants to receive information about the target OS.

CMF for a given OS starts,
The requesting OS is registered as an additional recipient for CMF data when it is active and a subsequent request is issued by an OS that has not yet registered to receive CMF data for the selected OS. Measurement activity for the selected OS is not affected.

Periodically, the IOP moves the data associated with a given OS to the CMB in the memory of each OS registered to receive data for this OS.

When the OS for a given OS is active and the SCM stop request is issued by either OS, the requesting OS is the OS registered to receive CMF data from the selected CMF. Removed from the list. As a result, the CMF for the selected OS
If the OS for receiving data is not registered,
The CMF for this OS is stopped as described above.

When the OS issues an SCM test request for selecting an arbitrary OS, only when the requesting OS is registered as a data receiving destination for the selected OS,
The stored result reflects the state of the CMF for the target OS. Otherwise, the response to the test request indicates that the CMF for the selected OS is down.

The acquisition of data by another OS uses the hypervisor privileges and security management mechanisms that exist in the prior art to provide security management whereby an operator or system administrator can use all other OSs. Allows each logical partition to be specified regardless of whether it is authorized to receive CMF data for. The OS that has the permission is another O
When the SCM start command for selecting S is issued, the processing is as described in this section. Otherwise, the start request is unsuccessful and the individual response code associated with this failure is stored in the command response block.

Normally, an OS running in a CEC partition is guaranteed by the CEC hypervisor that it does not need to know its own IID or that it is running in a logical partition environment. This allows each OS, if desired, to be executed as if it were the only OS in the CEC. The OS that is trying to recognize the partition of the other OS is describing the system configuration and using the CEC hypervisor's prior art services via files that can be used by the licensed OS itself. IID as well as other OS partitions can be obtained.

If the OS is trying to measure the I / O channel activity of multiple other OSs simultaneously, the requesting OS issues multiple SCM instructions, each SCM instruction "in the associated command request block (CRQB). OS selection IID "
Specify one of the other OSs in the field. A separate CMB is required for each such request.

[0046]

DESCRIPTION OF THE PREFERRED EMBODIMENT FIG. 1 shows a CEC with a central processing complex (CPC) partitioned between logical resource partitions 1-N. The hypervisor is located in partition 0, and OS-
1 to OS-N are arranged in sections 1 to N, respectively. Each compartment is dedicated to this, MS
-1 to MS-N, which has a portion of the CEC's main storage (MS) resources.

Hardware Microcoded Hypervisor Does Not Require Visible CEC Partition
Assigned to the CEC. If a software hypervisor were used for CEC instead, partition 0 would be assigned to the software hypervisor.

Input / output channel subsystem resources (channel resources) are not divided between OSs like CPUs. According to the invention of US Patent Application No. 898867,
All input / output resources (including channel resources) are different C
The OS in the PC partition can be shared, and the other of the I / O resources can be unshared by dedicating it to the allocated OS. US Patent Application No. 89
The 8867 example is incorporated herein by reference in describing the measurement mechanism herein.

Channel Time Measurement Logic FIG. 19 is a timing diagram representing the process performed by each circuit and microcode of the channel processor in the CEC for the preferred embodiment of the present invention.
There are up to 256 channel processors in the CEC. One channel processor for each channel is represented by a channel identifier (CHPID). The subchannel control block or shared subchannel control block in the I / O channel subsystem can specify up to eight CHPIDs available for the subchannel.

In FIG. 19, time T0 represents the start time of the operation by the channel processor.

The channel timing of FIG. 19 occurs regardless of whether this timing is used in the measurement. This timing is only used for CMF measurements when the SCM instruction is executed by the OS with the CRQB start command and the channel is used by this OS.

During the period from T0 to T1, the channel is idle (not in use) and no time accumulation is done by the channel processor.

The channel is for channel path communication,
Subchannel startup (SSCH) executed by some OS
When selected by an instruction or the like, time T1 occurs,
A timed period begins for the channel during which the channel processor measures usage time for the OS using it.

The channel processor has an internal clock which updates its clock time by one at some constant rate (eg, every 128 microseconds). The channel processor clock time is the absolute time value entered in the clock register (CPCR shown in FIG. 8) and is used in the channel timing process.

At time T1 (when the channel processor enters the period of use), the channel processor uses the absolute clock time from the clock register as the "start time register" (CPSTR shown in FIG. 8) provided by the processor. To store. When the channel processor has finished operating at time T2 and is in an idle (not in use) state, the channel processor provides the absolute clock time from the clock register, the "End Time Register (CPETR in Figure 8)". The channel processor then subtracts the value in the "start time register" from the value in the "end time register",
The subtraction of T2-T1 = “use time interval” with respect to the request source OS is performed. Then, the channel processor adds the obtained "use time interval" to the "stored use time" for each OS (shown in FIG. 9).

In order to correlate the accumulated usage time, if any, for each OS with the time interval in which the usage activity occurred, the channel if the OS has any activity during that time.・ Processor is each OS
Stores the "timestamp" value for. For example, every 2 seconds, the channel stores the contents of the Channel Processor Clock Register (CPCR) in a timestamp field for each OS that was in use during the previous 2 seconds (FIGS. 8 and 9). Shown in). In 2 seconds,
If there is no channel activity for the OS, the OS
The time stamp for is not updated.

Accumulated use time and time stamp (see FIG. 9)
(Shown in FIG. 2) both indicate the measurement set for the OS. This set is shown as the "Accumulation Time for OS-x" and "Time Stamp for OS-x" registers. A separate set of measurements is provided for the hypervisor and for each OS configured in the CEC that is configured to share the channel processor and is located in the local storage of the channel processor. ing.

FIG. 9 shows a "channel processor measurement block" (CHMB) provided in the local storage of each channel processor. The CHMB contains the accumulated usage time and time stamp value for each OS that uses the channel. There are up to 256 CHMBs in the CEC's channel subsystem, representing up to 256 physical channels supported by up to 256 corresponding channel processors present in the CEC.

The contents of each CHMB in each local storage of each CEC channel processor are periodically copied to the CEC's HSA Channel Image Measurement Block (CIMB) measurement set. For example,
Copied every 2 seconds.

Further, CE in the HSA of CEC is regularly
These measurement sets of C's CIMB are
Main memory channel measurement block (CM
B). For example, it is copied every 4 seconds.

In the hardware system area of CEC,
And these control blocks within the main storage of the respective OS logical partitions are described in detail later in this specification.

OS Channel Measurement Set (SCM) Command FIG. 2 shows an example of a “channel measurement set” (SCM) command, which is a program running under any OS (that OS). Must be located in the MS space). Using the SCM command in the OS program, start and stop the "channel measurement mechanism" (CMF) provided in each OS.
And test the situation. The mechanism is that all channels (shared or non-shared) configured for that OS for the period between the SCM start and SCM stop commands.
To measure the amount used. That is, measurements are taken for a time period controlled by executing the first SCM instruction that starts the mechanism and then executing the second SCM instruction that stops the mechanism. The preferred form of the SCM instruction requires that this instruction be executed only when its CPU is in a supervisory state.

Thus, the CEC hardware / microcode containing the I / O channel subsystem in which the SCM instruction is used by the OS program to construct a "channel measurement facility" for executing commands in response to measurement requests. The channel measurement command. OS except for limited functionality provided by special instructions configured for a particular purpose, such as SCM instructions.
The program does not have access to the CEC hardware / microcode configuration.

Although the form of one operand is shown in FIG. 2, many different formats can be arbitrarily given to the SCM instruction, such as by providing one, two or three operands. . Whatever instruction format is used, the SCM instruction has three different operands: the "command request block" (CRQB) in FIG. 3 and the "command response block" in FIG.
(CRPB), and "Channel measurement block" of FIG.
Place (CMB). Prior to executing the SCM command,
The CRQB, CRPB and CMB are arranged and configured by the OS program. The OS program is O
Information is placed in some of the fields in these blocks to be sent to the channel measurement mechanism, which is located in the hardware / microcode of the I / O channel that the S program cannot access. The format of one explicit operand and two implicit operands of the SCM instruction is shown in FIG. The first (explicit) operand places the command request block (CRQB) for the CPU.
The second (implicit) operand, the command response block (CRPB), is placed by the CPU immediately after the placement of the first operand, and the third (implicit) operand, the channel measurement block (CMB), is placed by the CPU. It is placed from the field in the 1 operand (CRQB). Therefore, the explicit operand one form of the SCM instruction places a block that places the other implicit operand, which is the form used in the IBM S / 390 architecture.

The SCM instruction, which has a start command in its CRQB command code, initiates the operation of the Channel Measurement Facility (CMF) for the specified OS. Each O
The Channel Measurement Facility (CMF) for S includes a channel processor, an input / output processor (IOP) that drives the channel processor, and associated control blocks,
It is included in the HSA (hardware system area) of the CEC storage device that the OS cannot access.

When started, the CMF collects information about the usage of the channel image in the local storage of that channel processor. This information is sent periodically to the HSA's control block to the channel image measurement block (CIMB) (FIG. 10). Then CIMB
This information in the microcode controlled CMF,
Microblock accessible control block (CIM
From B), it is periodically sent to the control block (channel measurement block (CMB) in FIG. 5) accessible by the OS. The OS program can access only this information in the control blocks accessible to the OS at any time.

When the channel measurement mechanism is started by executing the SCM instruction, the mechanism remains active until stopped by one of the following conditions:

1. Other SCM instructions are executed by the associated OS with the stop command shown in CRQB. Only the channel measurement mechanism for the specified OS is stopped. Channel measurement mechanisms for other OSs are not affected and can continue to operate. It is not necessary to specify the image number in non-hypervisor mode and in hardware hypervisor mode. In software hypervisor mode, the OS image number is the requesting OS if the SCM instruction is executed in interpreter format in passthrough mode.
To the CRQB by the hypervisor if the SCM instruction is not executed in interpreted form.

2. When the CEC power-on reset or the CEC system reset is performed, all channel measurement mechanisms for all OSs are stopped.

3. When the channel measurement mechanism enters the error stop state, only the channel measurement mechanism having the check stop state stops.

4. When the CEC or its channel subsystem goes into the check stop state, all channel measurement facilities (provided for all OSs) are stopped.

5. If the "OS image reset" function is executed by the hypervisor, the specified OS
Only the channel measurement mechanism for is stopped.

6. If a program check, protection check, uncorrected storage error, or uncorrected storage key error is detected by the channel subsystem when attempting to access the channel measurement block, the OS in which this condition occurred Only the associated channel measurement mechanism is stopped.

7. To recover a channel, one or more channels for which the channel subsystem needs to start an associated timer (used to calculate the channel usage time or channel time stamp): When an error condition is detected, only the channel measurement mechanism associated with the OS in which this condition occurred is stopped.

8. Depending on the hypervisor type,
If the system reset is done to the CEC as part of an automatically or manually initiated recovery function, then all channel measurement mechanisms for all OSs are stopped.

In this manner, the present invention selectively measures by the OS when using the input / output resources (whether shared or not) available to the program operating under the OS. Is possible.

OS command request block (CRQB) FIG. 3 shows the operand of the "channel measurement setting" instruction.
Indicates a "command request block" (CRQB) that specifies the type of command executed by this instruction. CR
QB is MS assigned to each OS in CEC
Is located in. The MS assigned by the OS partition is O
S, as well as all real storage addressable by all programs running under that OS. The OS (and its program) runs on other MS (C
(Assigned to other OS in EC) cannot be addressed.

The OS (or one of its programs) is C
Configure the RQB and give it to the CEC I / O channel subsystem as part of the channel measurement setup instructions,
The utilization of input / output resources (such as channels) used by each OS is measured asynchronously. The fields in CRQB shown in FIG. 3 are as follows.

L1: Specifies the length of the command request block such as 32 bytes.

Command Code (CC): Specifies a value that represents the type of operation of the Channel Measurement Facility to be performed. The meaning of each value is as follows. Start the 0 channel measurement mechanism. In hypervisor mode, the channel measurement mechanism is started only for the OS specified in the OS image ID field (if the OS selection IID field is zero, the channel measurement mechanism for other OS is not affected). . All fields defined in CRQB
Used when performing a starting operation. If the specified channel measurement mechanism is already active, the specified channel measurement mechanism remains active and normal instruction completion is indicated in the response block. In the latter case, the specified relocation zone, storage access
The key and channel measurement block address values are used to update all subsequent measurement blocks for the specified OS. Except when an error condition is detected by the channel subsystem for any of the specified values, in which case the measurement mechanism remains active and the previously specified value: Continue to use the value that was valid before the execution of the second start command. In non-hypervisor mode, it is not necessary to specify the OS image number (since only one OS is involved).

The 1-channel measurement mechanism is stopped. In hypervisor mode, both the command code field and the image identification field are used to perform the stop operation. The channel measurement mechanism for the specified OS is stopped, and no further access is made to the corresponding channel measurement block. Other channel measurement mechanisms for other OSs are not affected. If a stop command is specified and the specified channel measurement facility is not active, the specified channel measurement facility will remain stopped and normal command completion will be indicated in the response block. In non-hypervisor mode, only the command code field is used when performing the stop operation, and there is no need to specify the OS (since only one OS is involved).

Test the situation of the two channel measurement mechanism. The status of the channel measurement facility for the specified OS is returned to the requesting program in the command response block. In the hypervisor mode, the command code field and the CPC image identification field are used in performing the test operation. In non-hypervisor mode, only the opcode field is used in performing the test operation.

OS Image ID: In Hypervisor mode, this field contains the partition's IID (Image Identifier) code, which identifies the OS on which channel measurements are started, stopped, and tested. Note: In a CEC using a hypervisor, for security and transparency reasons, the OS does not access its OS image ID value, or the hypervisor does not
Since it is desirable not to be bothered by controlling C, the OS image ID value is set to the hypervisor.
If this is necessary, it is preferable to provide it. In this case, the OS program specifies all zeros in this field. All zeros are also specified when the CEC is operating in non-hypervisor mode with a single OS.

Relocation Zone: Relocation to access the required one of the MS's of multiple OS's containing the OS control block used by the channel measurement mechanism provided to the channel subsystem of the CEC. The value of the zone is used. If the CEC is not running in the hypervisor, the relocation zone will not be used.

Note: CEC using hypervisor
For security and transparency reasons,
It is not desirable for the hypervisor to access the value of the relocation zone, or to be bothered by the hypervisor controlling the CEC, so letting the hypervisor provide the zone relocation value when this is needed. Is preferred. In this case, the OS program specifies all zeros in this field. CEC is a non-hypervisor
When operating in mode (with a single OS)
All zeros are specified.

Key: Contains the storage access key needed to access the channel measurement block in the OS's main storage.

Channel Measurement Block Address Contains the address of the channel measurement block area in main memory for the requesting OS. The specified address is with respect to the origin of the associated relocation zone specified by the relocation zone field of this CRQB. C
If the EC is not controlled by the hypervisor,
The address becomes an absolute main storage address. This field is ignored if the operation code specifies a stop or test operation.

OS Command Response Block (CRPB) FIG. 4 is similar to CRQB, and "command response block" is arranged in any one or more of MS-1 to MS-N assigned to each OS in the CEC. Block "(CR
PB) is shown. OS (or one of its programs)
Provides space for CRPB immediately after CRQB in preparation for issuing channel measurement setup commands to the CEC channel subsystem. The channel subsystem inserts in CRPB a value that indicates how the requested operation was completed. The fields in CRQB shown in FIG. 4 are as follows.

L2: Designates the length of the command response block such as 8 bytes.

Response code: In order to execute the channel measurement setting instruction, a response code indicating the result of the request is sent to the channel.
Received from subsystem. Valid response is hex 00
01, 0003, 0004, 0102, 0103, and 0104.

Status: In hypervisor mode, the status of the channel measurement facility for the specified OS is provided by the channel subsystem. In non-hypervisor mode, the status of a single channel measurement facility is provided by the channel subsystem.
The meaning of each situation value is as follows.

The 0 channel measurement mechanism is active.

The 1-channel measurement mechanism is in a stopped state. The channel measurement mechanism has not been started, or has been stopped by performing a stop operation (OC = 1 in requesting CRQB).

A program check was encountered and stalled when the 2-channel measurement facility attempted to update the channel measurement block. The address of the channel measurement block is invalid. The address of the channel measurement block is invalid if it is not within the specified zone and at a storage location that is not configured.

3 When the Channel Measurement Facility tried to update the Channel Measurement Block, it encountered a protection check and is stuck. The storage access key specified in the command request block does not match the storage access key of the channel measurement block.

When the 4-channel measurement facility attempts to update the channel measurement block, it encounters an uncorrected storage error or storage key error and is stuck.

The 5-channel measurement mechanism has encountered an error and is in a stopped state. The condition that caused the error has been recovered. The facility can be restarted immediately by executing the set channel measurement command and specifying the start opcode.

The 6-channel measurement mechanism has encountered an error and is in a stopped state. The condition that caused the error has been recovered, but the channel measurement facility must not be restarted until a predetermined amount of time has elapsed. For example, do not restart the CMF unless the channel processor measurement clock is re-initialized and can be used again for time-of-use measurements.

7 The channel measurement mechanism has encountered an error and is in the check stop state. The channel subsystem cannot recover the condition that caused the error. External intervention is required to restart the mechanism.

Whether the state of the channel measurement mechanism is changed by a start operation or a stop operation, the input / output system reset function is executed, the OS image reset function is executed, or the specified channel measurement mechanism is executed. By repeating the test operation, the same status value can be obtained unless an error condition affecting the CRP is recognized.
B.

If the operation code of the requesting CRQB specifies a start or stop operation, the status field will be zero.

Channel Measurement Block (CMB) of OS FIG. 5 is similar to CRQB and CRPB and is arranged in any one or more of MS-1 to MS-N assigned to each OS in the CEC. Channel measurement block "(CMB).

Each OS has a "channel measurement setting" S for the I / O channel subsystem of the CEC executed by the CPU dispatched to perform the OS task.
In preparation for issuing a CM command, CMB (and
CRQB and CRPB) locations and a program to provide space for them.

In this preferred embodiment, I / O resource measurements are made periodically and asynchronously by each provided channel processor which makes time measurements on each channel and channel image. After the time measurement is sent to the measurement set in the control block HSA,
The time measurement of the HSA control block is C by IOP.
It is periodically copied to the control block of the MS of each OS in the partition allocated by each OS in the EC.

"Sample Count" in the CMB of FIG.
The field is commonly used for all measurement sets in the CMB. The operation of the sample count field is as follows.

Sample Count: Incremented once each time this CMB is updated by the Channel Subsystem.

Each CMB measurement set is a set of fields for each channel in the CEC, and in this example it is assumed to have 256 sharable channels. These channel sets are numbered from 0 to 255. Each channel set includes an "S" field, a "time of use" field, a "V" field, and a "time stamp" field for one channel while the channel is being used by the associated OS. There is. Other N O's in CEC
S may also have a similar CMB for the same channel in its respective MS.

The fields in each channel set of the CMB can only be activated if the associated channel is configured online in each OS. The operation of the fields within each channel set of the CMB is as follows.

V: "valid" field. If V = 1, the associated channel is configured to the requesting OS,
The usage time is accumulated and periodically stored in the channel usage time field. If V = 1, the associated "Channel Timestamp" field is also valid. V = 0
, The associated channel is not configured for the requesting OS and the contents of the associated channel usage time field and the time stamp field are not significant. The V bit is stored as zero if any of the following conditions exist:

1. The associated channel is not provided by CPC. 2. The associated channel does not provide channel usage measurements. 3. The associated channel is in a permanent error condition.

Channel Usage Time (CBT): Contains the accumulated channel usage time for each associated channel. A channel is in use if it is actively communicating with the control unit or device to which it is connected. CB if the associated channel is not shared
T is the sum of all channel usage periods. If the associated channel is shared by one or more other OSs, the CBT field contains the total channel duration for only the OS associated with the requesting channel measurement block. The accumulated value does not include the period of use for another OS sharing the channel. (Time interval is channel
Measured on the channel using the resolution appropriate for the type. ) Each channel is either another channel or a channel
Record channel usage time independent of image.

S: Shareable channel identifier field.
The S bit is significant only if the V bit is 1 (the item is valid). If S = 0 and V = 1, then the associated channel is configured for the target OS and the other OS
If S = 1 and V = 1, which cannot be shared by, the associated channel is configured for the target OS and can be shared by one or more OSs.

Channel Time Stamp (CTS):
Contains a timestamp value that identifies when the most recent channel usage time for the associated channel was recorded by the channel subsystem. Each time stamp applies only to the associated channel usage time of the same channel measurement set.

All fields of the channel measurement block are not significant until the measurement block has been asynchronously updated at least once by the channel subsystem.

HSA CPU-IOP communication block (C
ICB) FIG. 6 shows an HSA (hardware
"CPU-IOP communication block" (CIC) configured by hardware / microcode of system area)
B) is shown.

The fields of CICB are filled in by the corresponding fields of CRQB or by the SIE state description if the SCM instruction is executed interpretively. However, when the microcode fills the same labeled field in the CRPB, the status and response code field filled by the microcode are excluded.

HSA IOP Status Block (IOPS
B) FIG. 7 shows the "IOP status block" (I
The format of OPSB) is shown. IOPSB is CEC
There is an entry for each of the Channel Measurement Facility (CMF) in. A CMF can be provided in the hypervisor and for each of the N OSs in the CEC operating in hypervisor mode.

Each of the IOPSB items has a "status" field and a "channel measurement block address" field. The status field can have any of the following status codes:

0 = CMF is active for this OS 1 = CMF is down for this OS 2 = CMF is in the program check state, so this O
It has stopped for S. 3 = CMF is stopped for this OS because it is in a protection check state. 4 = CMF is a storage area or storage key error,
It has stopped for this OS. 5 = CMF has stopped for this OS due to CMF error. It can be restarted immediately by the program. 6 = CMF has stopped for this OS due to CMF error. The program must wait for a predetermined amount of CEC before restarting. For example, 16 seconds. 7 = CMF is in the check stop state for this OS.

Channel Processor Clock Register (CPC in Channel Processor Local Storage)
R), channel processor start time register (CPS
TR) and Channel Processor End Time Register (CPETR) Figure 8 shows the registers in the local storage of each of the channel processors in the CEC. One set of these three registers is provided for each physical channel controlled by each channel processor. These registers are used by each channel processor to continuously measure the period of use for each OS sharing the channel.

Channel Processor Measurement Block (CHMB) in Channel Processor Local Storage Figure 9
Is a channel processor measurement block (CHM in the local storage of each of the channel processors in the CEC).
B) is shown.

One CHM for each physical channel in the CEC
B is provided. Each CHMB has N measurement sets, one measurement set for each OS currently configured in the CEC. Each measurement set of CHMB stores information about one channel image (ie, physical channel usage by one OS). The fields in the CHMB measurement set are defined the same as the corresponding fields in the CMB measurement set.

Each channel processor updates its channel's accumulated usage time for each OS and stores it in the respective measured set of CHMBs in the processor's internal storage (its local storage).
The usage time for that channel is continuously measured. The present invention continuously measures and stores the channel image of each OS.

The continuous measurement and storage of the usage time for each OS channel image is not affected by the regular transfer of the accumulated time to the HSA, and the regular transfer has a statistical sampling effect. (As described in connection with FIG. 19).

Channel Image Measurement Information Block (CIMB) of HSA FIG. 10 shows a channel image measurement block (CIMB). CI in HSA for all channel images brought by CEC
One MB is provided. The CIMB has channel processor entries for all possible channel processors in the CEC, eg, 256 channel processors. Each channel processor item has N measurement sets, one measurement set for each OS image configured in the CEC. Each measurement set has the same fields as the measurement set in the channel processor measurement block (CHMB) of FIG.

Periodically (eg, every 2 seconds), the contents of all CHMBs (up to 256) are provided in the CIMB for all channel processors from all of their respective local storage. Is transferred to the corresponding item. That is, each item of CIMB is copied from each CHMB to one of the plurality of channel processors.

CIMB can only access CEC microcode and cannot be addressed by the OS.
Therefore, the contents of the CIMB are periodically (eg, every 4 seconds) transferred to the channel measurement block (CMB) in the storage of the OS, making this information addressable to the OS, which is the preferred embodiment. Is done in.

Normal Command Termination If the CEC is operating in hypervisor mode,
The response code 0001 indicates the normal completion of the SCM instruction given by the channel measurement mechanism to the specified OS after performing the operation requested by the CMF (the operation specified in the command request block for the SCM instruction). CMF has been started, stopped, or tested, as determined by the code and IID).

When the CEC is operating in a non-hypervisor mode, a response code of 0001 is sent to one OS of the CEC which was started, stopped or tested, as determined by the operation code specified in the command request block. Only indicates that the channel measurement mechanism is working. 000 in hypervisor mode
The response code of 1 is given only to the requesting OS indicated in the OS image ID field in the command request block of the SCM instruction executed by each OS. In hypervisor mode, any physical channel can be shared by multiple OSs in the CEC and only the use of the physical channel by each OS is reported to the OS. However, in hypervisor mode, it is not necessary for the physical channel to be shared by multiple OSs in the CEC, in which case the physical channel is used by only one OS and the measurement mechanism is responsible for the overall use of the physical channel. Report to OS.

Special Condition When a response code other than 0001h is stored in the command response block of the channel response block (CRPB) of FIG. 4, a special condition exists. Response codes other than 0001 indicate that instruction execution is suppressed for the OS executing the channel measurement setting instruction. In this case, the designated channel measurement mechanism is SCM.
It is left in the state it was in before the instruction was executed. SCM
The special condition instruction code placed in CRPB for the instruction is as follows.

^ 0003 ^: The response code 0003h is stored in the CRPB for the following reason.

1. The L1 field of CRQB contains a value other than a predetermined number of bytes, for example 32 bytes. 2. The fields of CRQB that are considered to be zero are not all zeros. 3. The CEC is not controlled by the hypervisor and the CRQB OS image ID is not zero. 4. The start operation is being performed by a hypervisor SIE instruction and the general conditions for interpreted execution are met, but the relocation zone field of CRQB is not zero. 5. Hypervisor SI for start, stop, or stop operation
It is executed by the E instruction and the general conditions for interpreted execution are met, but the IID field of CRQB is not zero. 6. The operation code of CRQB is not zero.

^ 0004 ^: When the SCM command is not given, the response code 0004h is stored in CRQB.

^ 0102 ^: Response code 0102h is stored in CRPB for the following reason.

1. A start operation is specified and the relocation zone
The field is not zero, but the specified zone is CE
Not given to C. 2. A start operation was specified, but the specified channel measurement block address is invalid. The measurement block address becomes invalid if any of the following conditions are recognized: -The address does not specify a proper location, such as a 4K byte boundary. -The address specifies a location that is not within the limits of the specified zone. -It specifies a position where the address is not configured. -The address specifies a location not given by the model.

^ 0103 ^: Response code 0103h is stored in CRPB for the following reason.

1. Storage access for the specified channel measurement block area with the specified storage access key
It does not match the key. If the specified storage access key of CRQB is zero, the channel measurement block key is not tested for a match value.

^ 0104 ^: Response code 0104h is stored in CRPB for the following reason.

1. The CPC is operating in hypervisor mode, but the specified OS image is not provided by the model or current I / O configuration.

Detailed Operation of SCM Instructions FIG. 11 is a flow chart showing the sequence of steps required for the CPU to execute a "set channel measurement" (SCM) instruction. First, in step 91, the CPU executing the instruction fetches the command request block (CRQB) and the command response block (CRPB) addressed by the instruction operand. CRQB contains the address of the OS Channel Measurement Block (CMB), which contains the implicit third operand of this instruction. C
The information inserted before the RQB is the length of the CRQB, the operation code, and the OS image I assigned to the OS partition.
D, the relocation zone number assigned to the OS partition, C
It contains the CMB address and the storage protection key used by the channel subsystem to access the MB.

OS image ID and zone relocation number are not pre-inserted into CRQB by OS, but SC
Inserted into the CRQB by the hypervisor if the M instruction is executed interpretively, or if the SCM instruction is executed interpretively and reissued by the hypervisor on behalf of the OS. Is preferred. In the preferred embodiment, the OS does not recognize the OS's assigned OS image ID or OS's relocation zone number, which is handled only by microcode and by the hypervisor.

In step 92, the CPU is the control block provided by the operand of the SCM instruction, S
Performs CM interpreter execution check and validity check.
The process of step 92 is expanded on FIG. 12 to determine if the CPU is running in SCM SIE interpreter execution mode, as well as the various plausibility checks previously described herein.

At step 93, the CPU determines CRQB.
Store the address in a control block that can only be accessed by the CEC hardware / microcode. C shown in FIG.
The PU-IOP communication block (CICB) is a part of the CEC memory that is accessible to the CEC microcode but inaccessible to the OS except by a limited method by special instructions that are given only to certain types of access (hereinafter , "Hardware system area" HSA).

In step 94, the CPU signals the input / output processor (IOP) of the CEC to continue execution of this SCM instruction, which is shown in FIG. This shifts the burden of performing SCM start, stop, and test operations from the CPU to the I / O channel subsystem. Step 95 then represents the CPU waiting for the IOP to complete the SCM instruction process.

When the IOP completes the SCM instruction process, the CPU enters step 96 of FIG. 11, step 9
End the CPU wait represented by 5. IO
The advantage of having P do this part of the execution is that the SCM instruction microcode is in the I / O channel subsystem, rather than in the CPU where the microcode requirements are stringent.

At step 96, the CPU executes the IOP.
The response code is fetched from the CPU-IOP communication block (CICB) generated during the processing of the SCM instruction. In step 97, the CPU stores the response code in the command response block (CRPB) of the MS of the requesting OS.
Step 97 occurs near the end of execution of the SCM instruction. The response code indicates to the OS program whether or not the execution of the SCM instruction was successful, and if not successful, the reason for the failure. These codes can have any of the values previously described herein.

In step 97, the CPU inspects the operation code to determine if a test operation has been performed, and if so, retrieves the CMF status from the CICB and stores it in the command response block. , C
Determine if step 99 is performed to indicate the status of the MF to the OS program.

FIGS. 13-15 show the channel subsystem process for performing start, stop, or test operations as part of the execution of SCM instructions. This starts from step 111, which starts from the OS storage area and uses the CRQB length, operation code, and OS of the source OS.
Retrieve the CRQB containing the image ID and relocation zone number, the channel measurement block storage access key, and the address of the channel measurement block.

Note: When the SCM instruction is executed by the hypervisor, CRQB contains only the OS image ID and relocation zone number. If the SCP instruction is executed interpretively by the OS, both of these values are passed to the IOP by the CPUSCM instruction microcode in the HSA's CICB. If the OS program is executing the SCM instruction interpretively, both the OS image ID field and the relocation zone number field are set to zero in the CRQB by the OS program before the CEC is controlled by the hypervisor. Only when not available does the OS program need to execute the SCM instruction. CPU microcode is O
The CICB ID, as well as the relocation zone number from the active SIE state description, as a means for providing these values to the IOP, as shown in step 1010 of FIG.
To copy.

Step 112 tests the CRQB opcode to determine if a start operation has been specified. If the operation is not start, the process proceeds to FIG. If start operation is specified, IOP is CE
Test the operating mode of C to determine if the hypervisor is controlling the CEC (step 11).
3). If no hypervisor is controlling the CEC, the process goes to step 116.

If the hypervisor is controlling the CEC (Yes path of step 113), the IOP tests to determine if the SCM instruction is executing interpretively in passthrough mode. (Step 114). Step 114 is the CPU of CICB
The SCM passthrough bit set by the SCM microcode is tested to determine if the SCM instruction is being executed interpretively under control of the SIE. CP in CICB, if running
Check the OS image ID and relocation zone number passed by U to determine if they specify the value given by CEC (step 11).
5). If these values are valid, the process proceeds to step 116.
Proceed to. If these values in the CICB are not valid, the appropriate error response code (described earlier in this specification) is sent to the HS.
It is stored in the command response code of the CICB of A (step 1112), and a signal indicating that the IOP executing the SCM instruction is completed is given to the waiting CPU (FIG. 15).
Step 135). The IOP then proceeds to look for other work tasks to perform, such as executor I / O requests for previously queued startup subchannel requests (step 136 in FIG. 15). If the CEC is not controlled by the hypervisor (test 114), these checks are not done and the process proceeds to step 116.

If the hypervisor is controlling the CEC, but the SCM instruction is not being executed interpretively in passthrough mode (no pass at step 114), the IOP performs the test of step 114A to
It is determined whether the OS image ID and relocation zone number designated by the hypervisor in each field of the RQB are valid. If these values are valid, the process proceeds to step 117. If these values are not properly specified by the hypervisor, the corresponding error response code is HSA CICB.
Stored in the command response code of
2) As described above, the CPU has the SCM in the IOP.
The signal that the execution of is completed is received.

Step 116 tests to see if the hypervisor is not controlling the CEC or the SCM.
Determines if the OS image ID and zone relocation number values in CRQB are properly designated as zero if the instruction is being executed interpretively. If the OS is properly assigned a value of zero, the process proceeds to step 117. If either of these values is non-zero, then the appropriate error code is stored in the HSA CICB command response code (step 1112) and the CPU returns the IO as described above.
A signal indicating that the execution of the SCM in P is completed is received.

Step 117 checks to determine if the CMB address specified by the OS is valid, and if all reserved fields in CRQB are properly specified with a zero value. If any of these checks fail, the appropriate error code is stored in the HSA CICB command response code (step 1112) and the CPU has finished executing the SCM in the IOP as described above. Receive a signal to that effect. If all of these checks are successful, processing continues at step 118.

Steps 118 and 119 perform a test to determine if the OS specified storage access key matches the OS's main storage CMB real storage access key. The IOP, using the specified key from CRQB, issues a fetch request to the storage address of the CMB and tests the result of the fetch operation. If the extraction is successful, the keys match and the process proceeds to step 1111. If the extraction is not successful,
The corresponding error code is stored in the command response code of the CICB (step 1112), and the CPU receives the signal indicating that the execution of the SCM in the IOP is completed, as described above. Note: If the CRQB key value is zero,
This allows the IOP to sequentially retrieve and store CMBs when the channel measurement information is stored in the CMB by the IOP on a regular basis, regardless of the actual key value of the CMB in storage. No testing is done.

Step 1111 performs the test and CE
Determine if C is running under control of the hypervisor. If not running, OS is 1
Only one operating in the CEC, the IOP stores an "activity" indication in the IOPSB entry of the HSA for OS image 1, and stores the storage access key value from CRQB and the CMB address from CRQB in OS image 1. Stored in the IOPSB item for. Hypervisor is CEC
If you are controlling multiple OSs, you can configure more than one OS, and the IOP will have an "activity" indication, a storage access key value,
And IOP of OS image with CMB address specified
Store in SB item. IOPSB status, key, and C
The MB address is then used by the IOP when it periodically updates each active channel measurement mechanism CMB for each respective OS. Upon completion of the IOPSB update, the process proceeds to step 134 of FIG. 15 where the IOP stores a normal completion response code in the CICB (step 134 of FIG. 15) to the waiting CPU.
As mentioned above, it signals that the SCM instruction has completed in the IOP.

FIG. 14 is similar to the check required by the IOP for a "stop" or "test" operation, and for a "start" operation as previously shown in FIG. Required hypervisor mode
IO for checking, SCM interpreter type execution check, and corresponding reserved field check
7 shows a P process. Check these (Step 12
2, 123, 123A, and 124) complete successfully, the opcode field of CRQB is checked for a "stop" operation (step 125). If the operation code does not specify the "stop" operation, the process is as shown in FIG.
Go to 5. If the operation code is determined to be a "stop" operation, the IOP checks to determine if the CEC is operating under the control of the hypervisor. If not running, only one OS is CEC
The CMF status for OS image 1 is set to "stop" in the IOPSB item of HSA for OS-1 (step 126A). CEC
Is operating under the control of the hypervisor, multiple OSs are running and the CMF status of the IOPSB item in the HSA for the specified OS image is set to the "stop" state (step 127). The process then proceeds to step 134 of Figure 15 to store the normal completion code in the CICB's command response code,
A signal indicating that the SCM instruction has been completed by IOP is issued to the waiting CPU.

FIG. 15 shows the IOP process for the "test" operation. In FIG. 15, if the "test operation" code is found in the operation code field of CRQB, the result of step 131 to step 132 is YES.
The pass is taken. If no "test" operation is found in the operation code field, an invalid SCP operation code is included in CRQB, the IOP stores the appropriate error response code in CICB (step 131B), and waits for the waiting CPU. , Signal that the SCM instruction has completed in the IOP, as described above.

Step 132 tests whether the CEC is using a hypervisor, and if so (Yes path to step 133), the channel measurement facility status for that OS is associated with the associated IOPSBO.
Copy from S item to CICB. The process then proceeds to step 134.

If it is determined by step 132 that the hypervisor is being used, then step 132
Enter A and input the status instruction existing in the OS-1 item to IO
Copy to PSB, then process goes to step 134.
Go to.

In step 134, a normal completion response code is stored in the CICB for the CPU microcode, and a signal indicating that the execution of the SCM instruction is completed by IOP is sent to the CPU which is waiting (step 135). The IOP then polls and executes other work tasks.

FIGS. 17 and 18 show the OS in the MS of each OS having an active channel measurement mechanism (CMF).
7 shows a timer driven routine for periodically updating the channel measurement block (CMB). For example, an IOP timer interrupt occurs every 4 seconds and inputs to step 151. The IOP then executes the process of FIG.
Step 151 checks the IOP status block (IOPSB in FIG. 7) for all CMFs. Step 151 checks if the IOPSB indicates that some CMF is active. If the CMF is not active, step 159 is entered and the IOP is released to do other work.

Step 152 checks if the CEC is using a hypervisor. If the hypervisor is not being used, step 152A is entered and a loop count of 1 is set. If the hypervisor is being used, then step 153 is entered and a loop count of N (which is the number of OSs configured in the CEC) is set. Step 154 then accesses the IOP status block (IOPSB) and also accesses the first status item for OS-1. The process of FIG. 18 is then executed to scan each channel processor entry (0-255) of the channel image measurement block (CIMB, see FIG. 10) to determine the currently specified OS (originally OS- Access the set of 2 words for that channel, and then set this set to the corresponding set of 2 words for that channel located in the channel measurement block (CMB) for that OS in that MS (see FIG. 5). Write.

Next, step 156 is entered, the loop count is decremented by one, and step 157 tests the resulting count to determine if it has gone to zero. If it is zero, then all 2-word sets are copied into the CMB (shown in step 158) for all OSs in the system and step 159.
At, the IOP is released to do other work.

FIG. 18 shows details of step 155 in FIG. If the CMF is not active, the process ends at point B in FIG. If any CMFs are active, the process proceeds to step 162, selects the first item of CIMB (FIG. 10) for channel 0, and proceeds to step 163.

Step 163 retrieves the set of two words containing the channel measurements for the selected OS and enters step 164.

Step 164 checks to determine if the next other channel entry will be processed by the CIMB. If yes, the process is step 163.
Return to and process the next item. If not, step 165 is entered and all 2 words retrieved are
Store the set in the CMB for each MS for the OS. Step 166 checks for any exceptions and if none exists, then FIG.
End at point B (this is step 15 in FIG. 17).
Go to 6). If there is any, step 166A is entered and the "stop" state is set for the CMF of the corresponding OS.
Stored in IOPSB for. This brings the CMF down to the associated OS and ensures that the IOP always encounters the same exception condition each time the IOP updates the CMB to the Active Channel Measurement Facility periodically.

[Brief description of drawings]

FIG. 1 shows a CEC with a CPU, I / O channel subsystem, and storage resources divided into logical partitions 1-N.

FIG. 2 illustrates a “Set Channel Measurement” (SCM) instruction.

Fig. 3 MS assigned to each OS of CEC
1 is a diagram showing a "command request block" (CRQB) arranged in any one or more of -1 to MS-N. FIG.

FIG. 4 is a diagram showing a “command response block” (CRPB) arranged in any one or more of MS-1 to MS-N assigned to each OS in the CEC, similar to CRQB. is there.

FIG. 5 shows a “Channel Measurement Block” (CMB) located in any one or more of MS-1 to MS-N assigned to each OS in the CEC, similar to CRQB and CRPB. It is a figure.

[FIG. 6] “CPU-IO” arranged in HSA (hardware system area) of CEC that cannot be accessed by OS
It is a figure which shows P communication block "(CICB).

FIG. 7: HSA showing the status of all channel measurement mechanisms
"IOP status block" (IOPS)
It is a figure which shows B).

FIG. 8 is a channel processor clock register (CPC), each provided with a channel processor.
R), channel processor start time register (CPS
TR) and the channel processor end time register (CPETR).

FIG. 9 shows a channel processor measurement block (CHMB) in the local storage of each channel processor.

FIG. 10: Channel image measurement block (CIM
It is a figure which shows B).

FIG. 11 is a flow chart showing a sequence of steps that result in an operation for a “set channel measurement” (SCM) instruction.

FIG. 12 is a flow chart showing the steps for performing a plausibility check to execute a “Set Channel Measurement” (SCM) instruction.

FIG. 13 is a flow diagram illustrating a sub-order of steps performed by an I / O processor to effect operations on a “set channel measurement” (SCM) instruction.

FIG. 14 is a flow diagram illustrating a sub-order of steps performed by an I / O processor to effect operations on a “set channel measurement” (SCM) instruction.

FIG. 15 is a flow diagram illustrating a sub-order of steps performed by an I / O processor to effect operations on a “set channel measurement” (SCM) instruction.

FIG. 16 is a flow diagram showing a sub-order of steps performed by an I / O processor to effect operations on a “set channel measurement” (SCM) instruction.

FIG. 17 is a flow chart showing the steps performed by the I / O processor to move the data in the channel measurement block from the HSA to the OS's partitioned main storage.

FIG. 18 is a flow chart showing the steps performed by the I / O processor to move the data in the channel measurement block from the HSA to the OS's partitioned main memory.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Steven Gardner Grassen, United States 12589, Sherwood Drive, Wallkill, NY 6 (72) Inventor Assaf Marron, United States 12603, Pawkeepsey, NY, Hollow Lane 1 ( 72) Inventor Kenneth James Oaks, United States 12590, Wappingers Falls, NY, Farm View Road 13 (72) Inventor David Emmet Stacky, United States 12603, Pawkeepsie, NY, Fox Run 123 (72) ) Inventor Leslie Wood Wyman Had, Parkekeepsie, NY 12601, USA Down Harbor Drive 1011

Claims (19)

[Claims] 1. Measuring the utilization of time by each input / output channel controlled by a channel processor operating under the control of a control program in which a measurement of the utilization of the input / output channel is made. A logical channel measurement mechanism (CMF) is configured in the input / output channel subsystem of the computer complex (CEC) to measure the channel utilization detected by the channel processor for the input / output channels used by the control program during the measurement period. Channel usage time consisting of accumulating each time segment by CMF and detecting each time segment by a channel processor measuring the time from the beginning of continuous use of the channel to the control program to the end of continuous use of the channel How to measure. 2. The process for determining channel usage time, further comprising the step of measuring by the channel processor each time segment from the start of usage status to the end of the usage time for the channel. To measure the channel usage time of the. 3. The method further comprises the step of excluding, from the step of accumulating each time segment used by any channel, the time of use of the channel for any other control program sharing this channel. 2. A method for measuring channel use time according to 2. 4. The method of claim 3, further comprising placing the control program in a logical partition of a CEC having a plurality of logical partitions, each partition containing an operating system with a control program. How to measure channel usage time. 5. A CEC in an I / O channel subsystem of a CEC is assigned to a CEC by permanently assigning a control program identifier to each of the CMFs and associating each CMF with a respective control program. 5. The method of measuring channel uptime according to claim 4, further comprising the step of associating with each of the plurality of control programs. 6. A central within the CEC for issuing a start command to an active CMF to initiate a measurement period in which the CMF is operating asynchronously with the CPU that is released to execute other instructions. CMF by arithmetic unit (CPU)
The method of measuring channel uptime according to claim 5, further comprising the step of: 7. The channel uptime of claim 5 further including the step of executing the CMF by a central processing unit (CPU) in the CEC to issue a test command to test the operating state of the CMF. How to measure. 8. Dedicated I / O channel processor to control each I / O channel, continuously measuring each time segment by each channel processor while the channel is in use, and using the channel. The method of claim 1, further comprising the step of providing time as channel utilization time. 9. The method of claim 8, further comprising executing a CMF start command only if a previously executed CMF test command indicates that the CMF is in a stopped state and not in an error state. To measure the channel usage time of the. 10. The method further comprises configuring each of the plurality of CMFs to share a channel processor in the CEC for measuring time segments for a plurality of control programs operating simultaneously in the CEC. A method for measuring channel usage time according to claim 5. 11. A CMF is identified for making measurements in a measurement command request control block (CRQB) located by a CMF instruction executed by a control program having the same identifier as that assigned to the CMF. The channel of claim 5, further comprising: responding by the CMF to a control program executing a CMF instruction by providing a measure of I / O channel utilization to a control block assigned to the CMF. How to measure usage time. 12. Operate each CMF independently of other CMFs, and for a utilized I / O channel selected on an OS basis, the shared channel contains only a proportionate amount of time each OS is in use. The method of measuring channel usage time according to claim 5, further comprising the step of enabling measurement of channel usage time. 13. Sharing a plurality of I / O channels between OSs, operating each CMF to an associated OS, and utilizing the shared I / O channel only when the channel is used by the associated OS. The method of measuring channel uptime according to claim 5, further comprising the step of measuring time. 14. Recording the start and end times for each time segment in which the channel is used by recording the current clock value of the clock as a time stamp entry in the local storage of each channel processor. The method of measuring channel uptime according to claim 1, further comprising. 15. At the end of each time segment, the start time is subtracted from the end time and the calculated time segment is added to the current contents of the accumulated use time field to obtain the accumulated use time for the channel processor. The method of measuring channel uptime according to claim 14, further comprising the step of calculating each time segment. 16. In a CEC containing multiple CPUs, occupying different resource partitions, each having an allocated portion of CEC main storage (MS) used by the occupying OS. A computer complex (CEC) having an I / O channel subsystem that supports channels sharable by multiple operating systems (OS) and a control unit for each I / O channel subsystem. A plurality of channel processors each of which senses the amount of continuous use time for each channel as a time segment of the measured value for the channel, and the time segments for the channels are summed and the sum is Means for recording in the channel storage item in the local storage area of the Means for recording a time stamp indicating the time to the occurrence of the end of the last segment provided in the CEC, and a plurality of channel measurement facilities (CMF) provided in the I / O channel subsystem storage of the CEC, Each C
MFs are assigned to perform I / O channel utilization measurements for each OS, and each CMF uses an I / O channel processor to control the I / O channels used by each OS. C
The contents of the MF and channel store and time stamp entries from all channel processor local storage, C
Periodically transfer to the corresponding item of MF to get the set of items for all channels and all CMFs,
The content of each CMF item is the accumulated usage time and time since the last reset of all CMF item contents.
A means for representing a stamp, and a CPU for executing the CMF command to any OS to start the identified CMF, wherein the identified CM
Since F starts the measurement period from the execution time of CMF, C
It has all the items reset by the start command issued by the MF command and stops the measurement period after the contents of the CMF have been obtained as a measure of the channel utilization by the OS, so that the CPU is Means to measure channel usage time consisting of CPU that executes CMF instruction. 17. The means for measuring channel uptime of claim 16 further comprising an identified CMF that is a CMF associated with an OS executing CMF instructions. 18. Means for measuring channel uptime according to claim 16, further comprising an identified CMF that is a CMF other than the CMF associated with the OS executing the CMF instruction. 19. A CPU that executes a CMF instruction to any OS to test the status of the identified CMF.
A means in the input / output processor for recording an identifier for each OS that executes a CMF instruction for the identified CMF, and a command response block (CRPB) arranged in the main storage area at a position designated by the CMF instruction. Means for responding to the OS by indicating the status of the identified CMF in the I / O processor.